You are on page 1of 19

Bulletin of Earthquake Engineering

https://doi.org/10.1007/s10518-021-01074-4

ORIGINAL ARTICLE

Wave velocities depending on shear strain, directionality,


and excess pore water pressure from wildlife liquefaction
array

Tadahiro Kishida1 · Chi‑Chin Tsai2 

Received: 6 July 2020 / Accepted: 4 March 2021


© The Author(s), under exclusive licence to Springer Nature B.V. 2021

Abstract
Data from the Wildlife Liquefaction Array, which is located at southern California’s Impe-
rial Valley in the USA, are analyzed. Variations in shear wave (­ Vs) and compression wave
­(Vp) velocities are identified during strong motions. Shear strains (γ) are computed from
recorded acceleration time series between the sensors. All these variables are compared
with the ground-surface accelerations and excess pore water pressure ratio ­(Ru) at differ-
ent depths. Results show that R ­ u starts to accumulate when γ exceeds the threshold strain
of 0.01–0.015%. V ­ s is negatively correlated to ­Ru and γ. Further investigation shows
that ­Vs degradation depends on maximum horizontal shear strain with the different azi-
muthal angles at the same time window. By contrast, V ­ s reduction with R
­ u is caused by
the decrease in effective stress; hence ­Ru reduces ­Vs regardless of the azimuthal angles. ­Vp
degradation is limited compared with that of V ­ s under strong shaking, indicating that the
constrain modulus is determined by the combination of soil skeleton and pore fluid.

Keywords  Downhole array · Wave velocity · Shear strain · Excess pore water pressure ·
Sand · Seismic loading

1 Introduction

Site response analyses compute the amplifications of seismic waves propagating through
subsurface soil layers to account for local site effects. Field observations have shown that
the development of excess pore water pressure influences the seismic wave amplifica-
tions and frequencies of ground motions (Zorapapel and Vucetic 1994). Therefore, effec-
tive stress site response analysis that include the generation of excess pore water pressure

* Chi‑Chin Tsai
tsaicc@nchu.edu.tw
Tadahiro Kishida
tadahiro.kishida@ku.ac.ae
1
Department of Civil Infrastructure and Environmental Engineering, Khalifa University of Science
and Technology, PO Box 127788, Abu Dhabi, UAE
2
Department of Civil Engineering, National Chung Hsing University, 145 Xingda Road,
Taichung 40227, Taiwan

13
Vol.:(0123456789)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

and associated cyclic softening of soil under strong shaking (e.g. DESRA-2, Lee and Finn
1978; DESRAMOD, Vucetic and Dobry 1986; D-MOD, Matasovic and Vucetic 1995;
DEEPSOIL, Hashash et  al. 2020) are believed to better predict site responses. However,
numerical prediction of the wave propagation Tsai et  al. (2017), generation and redistri-
bution of excess pore water pressures (Thilakaratne and Vucetic 1990), and the resulting
vertical settlement remains challenging even under simplistic level-ground free-field con-
ditions. Kramer et al. (2011) listed the challenges in validating 1D site response analyses
using vertical array data. One issue was “directionality”: the combinations of two horizon-
tal and vertical ground motions affect the generation of excess pore pressure in the soils
(e.g. Ishihara and Yamazaki 1980). However, most site response analyses consider only
a single component. If the site response analyses are performed with different horizontal
components, then pore water pressure are generated with varying levels, and the results of
site response analysis are influenced consequently. On this basis, a theoretical limitation
exists in current site response codes to compute horizontal and vertical wave propagations.
Dynamic site response properties and soil behaviors from the downhole data have been
investigated through different techniques. Wave velocities were calculated during strong
motions by computing the wave travel times between sensors (Elgamal et al. 1995; Haddadi
and Kawakami 1998). Inversion analyses were performed in frequency domain (Agu-
irre and Irikura 1997) and time domain (Glaser and Baise 2000) to identify dynamic site
response properties varying with time. Furthermore, stress–strain behaviors were extracted
to understand soil nonlinear behaviors during strong shakings (Zeghal and Elgamal 1994;
Zeghal et  al. 1995; Tsai and Hashash 2008,2009). Although many methods have been
developed to analyze downhole data, the variation of dynamic site response properties and
nonlinear soil behaviors with the development of pore water pressure was rarely investi-
gated (Zorapapel and Vucetic 1994; Groholski et al. 2014) because the measurements of
pore water pressures were not available for many downhole arrays in the last two decades.
Moreover, the directionality of dynamic site response properties during strong shakes was
rarely explored (Kishida et al. 2018). Therefore, this study analyzes the data from wildlife
liquefaction array containing measurements of acceleration and pore water pressure record-
ing. These data were previously analyzed and showed reasonable results on the recorded
variables (Steidl et al. 2014).
This study calculates shear wave velocity ­(Vs) and compression wave velocity ­(Vp) dur-
ing strong motions using Normalized Input–Output Minimization (NIOM) (Haddadi and
Kawakami 1998). Shear strains (γ) are also computed from the recorded acceleration time
series following the previous methods (Zehgal et al. 1995; Kishida et al. 2020). With the
use of these data with recorded excess pore water pressure, the dynamic site response prop-
erties were investigated by considering the generation of pore water pressure and the direc-
tionality of ground motions. The findings can help guide 1D effective stress site response
analyses to overcome the current practical limitations.

2 Wildlife liquefaction arrays and data analysis

The Wildlife Station was established by the US Geological survey in 1982 on a flood-
plain in the Imperial Valley of Southern California, wherein sand boils were developed
during the 1981 Westmorland earthquake (Bennett et  al. 1984). The station has been
operated under United States National Strong-Motion Network (NP) with Station ID
of 5210. In 2003–04, the Wildlife Liquefaction Array was instrumented as part of the

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

US National Science Foundation (NSF) Network for Earthquake Engineering Simula-


tion (NEES) (Youd et  al. 2007). The station has been operated under UC Santa Bar-
bara Engineering Seismology Network (SB) with Station ID of WLA. These stations are
separated by approximately 70 m.
Figure  1 shows the geological profile and instrumentation layout of Wildlife Liq-
uefaction Array (NEES@UCSB, http://nees.ucsb.edu/data-porta​l). The near-surface
soil layer is composed of a 2.5 to 3.0  m-thick silty clay to clayey silt. This layer is
underlain by silt, silty sand, and sandy silt layer with a thickness of 3.5–4.0  m and is
highly susceptible to soil liquefaction (Bierschwale and Stokoe 1984). A thick silty clay
to clay layer underlies the liquefiable sand layer. The instrumentation consists of tri-
axial accelerometers (Kinemetrics Episensor ES-T) and pressure transducers (Quanterra
330 L) installed at 9 and 10 different locations, respectively. The sensors have sampling
rates of 200  Hz. The accelerometers are located at ground surface, at depths of 2.5,
5.5, 7.7, 30, and 100 m. Locations 01 and 04 are just above and below the liquefiable
layer, and locations 02 and 03 are within the liquefiable soil. The pressure transducers
are placed at ground surface, at depths of 2.64, 2.95, 3.28, 3.51, 4.30, 4.42, 4.71, and
6.23  m. Locations 62/63 and 66/67 are installed at the top and the bottom of the liq-
uefiable layer, respectively, and locations 64/65 are at the middle of this layer. Table 1
lists the 13 earthquake events analyzed in this study. Peak ground acceleration (PGA)
and hypocenter distance range from 0.003 to 0.31  g and 6.1 to 350  km, respectively.
Moment magnitude (M) ranges from 3.4 to 7.2. The recorded data can be downloaded
from NEES@UCSB (http://nees.ucsb.edu/data-porta​l).
Figure 2 shows the ­Vs and ­Vp profiles by suspension logging method (NEES@UCSB,
http://nees.ucsb.edu/facil​ities​/wla). The liquefiable sand layer has the ­Vs of approxi-
mately 150 m/s. The ­Vp substantially increases below the depth of the 2.5 m, indicating
that the water table exists around this depth. Boring data showed that the water table
was located around 1.3-m depth (http://nees.ucsb.edu/facil​ities​/wla).

Fig. 1  Geological profile and instrumentation layout of the Wildlife Arrays; red and blue circles show
accelerometers and pore-water pressure sensors, respectively. Numbers indicate sensor location codes.
Depth of soil layers and soil types are shown from the ground level (NEES@UCSB, http://nees.ucsb.edu/
data-porta​l)

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Table 1  Analyzed earthquake events for wildlife liquefaction arrays


CI event ID Origin time (UTC) Magnitude (M) Hypocentral Component PGA (g)
distance (km) (NS, EW, UD)

14607652 2010-04-04T22:40:43 7.2 94.8 (0.103, 0.139, 0.056)


15199681 2012-08-26T19:31:22 5.4 6.1 (0.136, 0.168, 0.329)
15200401 2012-08-26T20:57:57 5.5 7.8 (0.183, 0.181, 0.202)
15200489 2012-08-26T21:15:29 4.1a 11.3 (0.120, 0.080, 0.043)
15201537 2012-08-26T23:33:25 4.6a 18.2 (0.274, 0.279, 0.079)
15202921 2012-08-27T04:41:36 4.5a 12.3 (0.255, 0.313, 0.168)
15203249 2012-08-27T06:31:27 3.4b 12.2 (0.089, 0.042, 0.039)
37298672 2014-12-24T05:51:51 4.2 13.2 (0.012, 0.017, 0.021)
37166079 2015-05-21T03:15:30 4.1b 17.6 (0.039, 0.015, 0.022)
37374687 2016-06-10T08:04:39 5.2 93.4 (0.013, 0.011, 0.003)
37644544 2016-07-31T16:21:06 4.0 26.6 (0.018, 0.011, 0.002)
38443183 2019-07-04T17:33:49 6.4 342.1 (0.003, 0.003, 0.001)
38457511 2019-07-06T03:19:53 7.1 352.6 (0.007, 0.012, 0.003)
a
­ b. bMagnitude is ­ML
 Magnitude is m

Fig. 2  Vs and ­Vp profiles at


Wildlife Liquefaction Array site
by suspension logging method
(http://nees.ucsb.edu/facil​ities​
/wla)

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

3 Data processing methodology

Fig. 3  Flow chart of data processing method to compute a wave velocities and b strain time series

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Fig. 3  (continued)

Three-component acceleration time series and pore water pressure measurements are
downloaded at the station of SB.WLA. From these recordings, the wave velocity and strain
time series are computed between downhole sensors. Figure 3a shows the flowchart of data
processing. Two-component horizontal acceleration time series were first rotated to the
specified azimuthal angles. Wave travel times between sensors were calculated by NIOM
as previously reported (Haddadi and Kawakami 1998). The NIOM minimizes the squared
Fourier amplitude spectra of the ground motions at the surface and at depth using Lagrange
multipliers method. The incident and reflected wave travel times are returned, whereas the
incident wave velocities are only selected because of the large uncertainties in the reflected
waves (Elgamal et  al. 1995; Mogi and Kawakami 2012; Kishida et  al. 2018). Figure  3b
shows the computation of the displacement time series by applying filters and baseline cor-
rections modified from the past studies (Chiou et al. 2008; Ancheta et al. 2013). Fourier
amplitude spectrum (FAS) and signal-to-noise ratio (SNR) of time series were reviewed

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

at different locations to select consistent high-pass corner frequencies (­ fc-hp) depending on


the earthquake. Low-pass corner frequencies are also selected by reviewing FAS and SNR
for each component. Acausal Butterworth filters are applied in the frequency domain to
remove low- and high-frequency noise. This process is consistent to Zehgal et al. (1995),
but different from Kishida et al. (2020) in which the ­fc-hp are selected for each component.
This adjustment is important to compute strain time series when M > 6 because the result-
ing strain amplitudes increased substantially due to the small difference in f­c-hp between
components and locations. The ­fc-hp was selected between 0.5 and 0.6 Hz when M > 6 to
remove the influence of surface waves. When the displacement time series are obtained by
the double integration of the acceleration time series, the strain time series are computed
by subtracting displacement time series divided by the separation distances between sen-
sors. The method is adapted from the previous studies on filter type and mathematical for-
mula (Zehgal et al. 1995).

4 Typical trends of observed data

Figure 4 shows time series of event 15201537 in Table 1, (a) acceleration at ground surface
and at the depth of 7.7  m in north–south (NS) direction, (b) excess pore-water pressure
ratio ­(Ru), defined as the ratio of excess pore-water pressure to effective vertical stress
before strong shaking, (c) maximum shear strain amplitude (γ), and (d) identified V ­s
between ground surface and sensors at depth. ­Vs is the average value from the different
azimuthal angles. γ is computed by square root√of sum of square (SRSS) for NS, east–west
(EW) and up–down (Z) components (i.e.,= 𝛾xz 2 + 𝛾 2 + 𝜖2  
yz z
). The ­Ru sharply increases
with S-wave arrival at 15 s, at which the γ also increases from 0.02 to 0.2% (Fig. 4a, b, c).
The developed excess pore water pressures are redistributed within the liquefiable sand
layer after the strong shaking. Simultaneously, the high ­Ru (e.g., location 63) decreases,
and the low R ­ u (e.g., location 62, 64–67) increases. R
­ u variation of the different locations
decreases with time (from 15 to 100 s in Fig. 4b). The largest γ is recorded between 2.5 and
5.5  m (i.e., location 01–02) with relatively low frequency after strong shaking (red line
from 17 to 31 s in Fig. 4c), and the largest ­Ru is observed at a depth of 3.51 m (location 63
in Fig. 4b). The obtained V ­ s between ground surface to the depth of 7.7 m is approximately
140 m/s prior to the strong shake (Fig. 4d), which is consistent with the field testing (Bier-
schwale and Stokoe 1984) and the observation (Fig. 2). The V ­ s decreases to 100 m/s at 15 s
with large γ and the generation of excess pore water pressure. After the S-wave packet, the
­Vs gradually recovers when ­Ru and γ decrease with time, indicating a negative correlation
between these variables. The slow recovery rate of V ­ s might be due to the clay layer near
ground surface that prevents the dissipation of excess pore water pressure to the ground
surface.
Similar observations are found in the 15202921 event (Fig.  5). The V ­ s decreases to
90 m/s at 16 s, and the γ increases from 0.02 to 0.2% with the arrival of S-wave (Fig. 5a,
b, c). After the S-wave packet, ­Ru is redistributed within the liquefiable layer (Fig. 5b). ­Vs
also remains as low as 120 m/s at 50 s (Fig. 5d), which is consistent with that in Fig. 4. The
­Ru is larger in event 15202921 (Fig. 5b) than in event 15201537 (Fig. 4b). Similar phenom-
enon is observed for γ (Figs. 5c, 4c). The ­Vs is smaller in event 15202921 (Fig. 5d) than in
event 15201537 (Fig. 4d). These observations in Figs. 4 and 5 show that γ, ­Ru, and ­Vs are
correlated within and between different earthquake events. The correlations among these
variables are further investigated by reviewing these data in the next section.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Fig. 4  Recorded time series in event 15201537 a NS-component acceleration at ground surface (00) and
depth = 7.7  m (04), b ­Ru within a liquefiable layer at different locations, c γ between two locations, d ­Vs
between ground surface and depth of 5.5 m (00–02) and ground surface and depth of 7.7 m (00–04)

5 Correlations among PGA, γmax, ­(Ru)max, and ­(Vs)min

The correlations among PGA, γmax, ­(Ru)max, and (­ Vs)min are reviewed from different earth-
quake events. PGA is defined by RotD100, in which the maximum value is selected over
all rotation angles from two horizontal components of motions (Boore 2010). γmax, ­(Ru)max,
and ­(Vs)min are defined as the maximum and minimum values of these variables from the
entire time series in the event. Figure 6a shows the correlation between PGA and γmax. γmax
increases with PGA, indicating a clear positive correlation between these variables. γmax
reaches to 0.12% when PGA becomes greater than 0.3 g. Figure 6b shows that the (­ Ru)max
at location 62 increases with PGA. The ­(Ru)max reaches to 0.15 when the PGA becomes
greater than 0.3  g. By comparing these figures, large M earthquakes (M > 6) generates
higher ­(Ru)max (Fig.  6b) compared with the smaller M even though γmax and PGA cor-
relate well for the entire M range. This finding could be due to the accumulation of strain-
induced damages caused by the long duration of ground motion with M > 6.
Figure 7a shows the data for γmax and (­ Ru)max. These variables are positively correlated,
in which ­(Ru)max reaches to 0.4 when γmax becomes 0.12%. ­Ru starts to increase when γmax

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Fig. 5  Recorded time series in event 15202921 a NS-component acceleration at ground surface (00) and
depth = 7.7 m (04), b ­Ru within a liquefiable layer at different locations, c γ between two locations, d Vs
between ground surface and depth of 5.5 m (00–02) and ground surface and depth of 7.7 m (00–04)

(a) (b)

Fig. 6  Correlations between PGA, γmax, and ­(Ru)max from 13 earthquake events in Table 1. PGA is obtained
at Location 00. γmax is obtained between Location 00 and 04. ­(Ru)max is obtained at Location 62. a PGA
versus γmax, b PGA versus ­(Ru)max

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

(a) (b)

(c)

Fig. 7  Correlations between γmax, ­(Ru)max and ­(Vs)min from 13 earthquake events in Table 1. The γmax and
­(Vs)min are obtained between Location 00 and 04. (­ Ru)max is obtained at Location 62. a γmax versus (­ Ru)max,
b γmax versus (­ Vs)min, c ­(Ru)max versus (­ Vs)min

exceeds approximately 0.01%. This γmax would be defined as the cyclic threshold shear
strain (γtv) for pore water pressure buildup (Dobry et al. 1982). Vucetic and Dobry (1986,
1988) conducted the cyclic testing on Wildlife sands and reported that γtv = 0.017%. Fig-
ure 7b shows the data between ­(Vs)min and γmax. These variables are negatively correlated
because ­(Vs)min decreases to 90 m/s and γmax increases to 0.12%. The ­Vs decreases substan-
tially when the γmax exceeds γtv ≈ 0.01%. This observation is mainly due to the develop-
ment of excess pore water pressure. This influence on seismic-wave amplifications was dis-
cussed based on the field observations (Zorapapel and Vucetic 1994). Figure 7c shows the
data for ­(Ru)max versus ­(Vs)min. These variables are negatively correlated for all M values
where ­(Vs)min decreases to 90 m/s and ­(Ru)max increases to 0.4. Figure 7c shows the data, in
which ­(Vs)min decreases from 140 to 128 m/s, although the ­(Ru)max is less than 0.05. Never-
theless, these data points are well correlated between γmax and ­(Vs)min in Fig. 7b. Therefore,
­(Vs)min is primarily influenced by γmax when (­ Ru)max is less than 0.05; however, the increase
in γmax and ­(Ru)max influences (­ Vs)min when (­ Ru)max is greater than 0.05.
Figure  8 shows the correlation between γmax and ­ (Ru)max where γmax is computed
between locations 01 and 02. ­(Ru)max increases with γmax and starts to increase signifi-
cantly when γmax is greater than 0.01–0.015%. By contrast, ­(Ru)max is negligible when γmax
is smaller than 0.01–0.015%. This observation indicates that γtv exists within these strain
levels. The manifestation of γtv from the observed data at the Wildlife Station was dis-
cussed by Matasovic and Vucetic (1996). The observation shows that the γtv of 0.017%

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Fig. 8  Correlations between
γmax versus (­ Ru)max. The γmax is
computed between Location 01
and 02. The (­ Ru)max is obtained
at Locations 62–65 within the
liquefiable layer

obtained by the laboratory experiments (Vucetic and Dobry 1986) reasonably matched to
the field observation during 1987 Superstition Hills Earthquake. Thilakaratne and Vucetic
(1990) showed that the onset of the pore water pressure buildup was reasonably predicted
by seismic response analysis by DESRAMOD at Wildlife Station by using γtv = 0.02% as
an input variable. Hsu and Vucetic (2004) reported that γtv ranges from 0.01 to 0.02% for
sand. Therefore, the field observations in this study are reasonably consistent with previous
laboratory studies, data analyses, and numerical evaluations.

6 Variation in ­Vs along different azimuthal angles

Figure 9a and b show the observed ­Vs and γaz with different azimuthal angles during event
15201537 at the time window of 14–16 and 60–62 s, respectively. γaz denotes the maxi-
mum horizontal shear strain in different azimuth within the time window where V ­ s was

Fig. 9  Vs and γaz in different azimuth angle during and after development of power water pressure in event
15201537. Time window from a 14 to 16 s, where the ­Ru increases from 0 to 0.2, b 60–62 s, where the ­Ru
is stable about 0.25

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

computed (Fig.  3a). Figure  9a shows the results when R ­ u increases from 0 to 0.2 at the
arrival of S-wave packet. Under different azimuthal angles, γaz and ­Vs range from 0.04
to 0.1% and 80 to 120 m/s, respectively. ­Vs values vary depending on γaz in different azi-
muthal angles. V­ s is low when γaz is large, thus indicating their negative correlations. Simi-
lar observations were previously reported from downhole array data at clayey site (Kishida
et al. 2018). Figure 9b shows the results when R ­ u is stable at 0.25. γaz and V­ s range from
0.0014 to 0.0019% and 120 to 130 m/s, respectively. Although γaz is small, V ­ s is smaller
than 140 m/s of the pre-shaking condition in Figs. 4 and 5. This finding indicates that V ­ s
decreases due to the developed ­Ru of 0.25. ­Vs does not range with different azimuthal
angles because of the small variation in γaz in Fig. 9b. As displayed in Fig. 9a and b, V ­ s is
controlled by the combination of γaz and R ­ u with different azimuthal angles, in which the
influence of γaz is orientation-dependent, whereas that of ­Ru is orientation-independent.
Figure 10 shows the ­Vs and γaz with different ­Ru from the events with M < 6.0 computed
at different time windows with different azimuthal angles. The presenting data are after
smoothing within ± 1.0 s and ± 30°. The figure only shows the data with γaz > 0.001%. ­Vs
consistently decreases with γaz even with varying ­Ru values. Moreover, V ­ s shows a decreas-
ing trend when R ­ u increases under the same γaz. These results show that given the devel-
oped ­Ru, ­Vs and γaz have a unique relationship that is parallel with different ­Ru values. Sim-
ilar trends were observed from another downhole data (Tsai and Hashash 2009) presented
with number of cycles instead of ­Ru based on the analysis with a single motion component.
These observations are obtained with R ­ u < 0.6 and should not be extended to higher R ­u
values.
Figure 11 shows the ­Ru and shear strain with different definitions. The data are extracted
when ­Ru increases during S-wave packets. The shear strain and R ­ u are obtained between
locations 01 and 02 and at 63, respectively. Figure  11a shows the data between ­Ru and
γ. No correlation is observed because the pore water pressure is induced after the devel-
opment of γ. Figure  11b shows the data between R ­ u and γcum.max. γcum.max is defined as

Fig. 10  Vs and γaz with different


­Ru from the events with M < 6.0
in Table 1. The V
­ s and γaz are
computed in different azimuthal
angles between Location 00
and 04

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering


Fig. 11  Ru vs. shear strain by different definitions a γ defined by 𝛾xz
2 + 𝛾 2 + 𝜖 2  , b maximum of the γ from
yz z
the beginning to the current time (γcum.max), c maximum of the γaz in NS component from the beginning to
the current time (γNS.cum.max), d maximum of the γaz in EW component from the beginning to the current
time (γEW.cum.max). The R­ u and γ are computed at Location 63 and between Location 01 and 02, respectively

maximum of the γ from the beginning to the time at which the ­Ru is observed. These vari-
ables exhibit a correlation, in which ­Ru starts to increase when γcum.max exceeds 0.015%.
This finding is similar to Fig.  8, but Fig.  11b plots multiple points from one earthquake
event. Figure  11c and d plot ­Ru with γNS.cum.max and γEW.cum.max, respectively, which are
defined as the maximum of γaz in NS and EW directions, respectively, from the begin-
ning to the time at which the ­Ru was observed. Similar to Fig. 11b, both figures show clear
correlations between these variables, in which ­Ru starts to increase when γNS.cum.max and
γEW.cum.max exceed 0.013% and 0.018%, respectively. From Fig. 11b, c, and d, no clear dif-
ference in correlations is found between these variables probably because of the limited
directionality of the analyzed time series.
Byrne (1991) reported that the increment of pore water pressure is related to a combina-
tion of shear strain amplitude and plastic volumetric strain prior to load application follow-
ing the study by Martin et  al. (1975). Strain-based pore-pressure models have also been
proposed (Vucetic and Dobry 1986, 1988; Matasovic and Vucetic 1993, 1995). From the
definition of shear strain amplitude, γcum.max might have a better correlation with R ­ u than
γNS.cum.max or γEW.cum.max and thus is suggested to be used in the strain-based model. The
accumulation of pore water pressure under directional loading was discussed by Vucetic
and Dobry (1986, 1988) to incorporate its effect into the cyclic pore water pressure model.
Reyes et al. (2019) shows that the bidirectional shaking produces 60% higher ­Ru compared
with the unidirectional shaking using soil constitutive models. The multiple-direction shak-
­ u by bidirectional shaking to unidirectional shak-
ing table test also shows that the ratio of R
ing is up to 3–5 (Ueng and Lee 2015). On the basis of these studies, practical solutions
might be developed to implement pore water pressure for 1D site response analyses by con-
sidering directional ground motions. However, in current practice, no validated approach

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

has been established to resolve this issue. As summarized from the downhole observation
in this study, the development of ­Ru is related to γ, and ­Vs ranges with the azimuthal angles
as a result of γaz and ­Ru combination. These aspects must be considered when modeling
soil behaviors coupled with pore water pressure developing in site response analyses for
propagating horizontal motions.

7 Variation in ­Vp and ­Vs under strong motions

Figures 12 and 13 show the variation in ­Vp and ­Vs between locations 00 and 05 for events
15201537 and 15202921, respectively. The developed ­Ru during these events is approxi-
mately 0.4–0.6 in maximum from Figs. 4 and 5. Acceleration time series at ground surface
are also presented. ­Vp and ­Vs are smoothed within 1.0  s, where ­Vs is the average from
different azimuthal angles. The separation distance between sensors is 30  m. When the
selected separation distances are shorter than 30 m, the obtained V­ p is unstable because of
the limited resolution based on the sampling frequencies of the sensors and high ­Vp value.
These observations show that V ­ s decreases due to the large γ and pore water pressure
during strong shakes (Figs. 12c, 13c); however, the reduction of ­Vp is limited (Figs. 12b,
13b). For instance, ­Vs drops from 220 to 170 m/s by 23% (Fig. 12c) but ­Vp decreases from
1140 to 1050 m/s by 9% (Fig. 12b). This finding indicates that the apparent constrain mod-
ulus from ground surface to 30 m depth remains approximately linear even when the shear

Fig. 12  Variation in wave velocities during event 15201537 a acceleration time series at ground surface in
NS direction b ­Vp between ground surface (00) and depth = 30 m (05), c ­Vs between ground surface (00)
and depth = 30 m (05)

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Fig. 13  Variation in wave velocities during event 15202921 a. acceleration time series at ground surface in
NS direction b ­Vp between ground surface (00) and depth = 30 m (05), c ­Vs between ground surface (00)
and depth = 30 m (05)

modulus (G) exhibits a high nonlinearity (G/Gmax = 0.6). Similar trends are observed in
Fig.  13. ­Vs drops from 220 to 160  m/s by 27% (Fig.  13c) and V ­ p drops from 1,110 to
975 m/s by 12% (Fig. 13b).
The differences in soil behavior in vertical and horizontal direction are explained as
follows (Tsai and Liu 2017; Liu and Tsai 2018). In the horizontal direction, shear waves
are propagated through the soil skeleton but not through the water. Therefore, the ­Vs are
influenced by the strain amplitude and the softening induced by the developed excess pore
water pressure. By contrast, compression waves in the vertical direction are propagated
through the soil skeleton and the water. When the soil is saturated, the compression waves
are mainly propagated by the water because its bulk modulus is generally higher than that
of soil. When the soil is unsaturated, the compression waves are propagated by the soil
skeleton. Therefore, the ­Vp are influenced by the strain amplitude when the soil is unsatu-
rated. Figure  13 shows that the V­ p recovered very quickly after strong shakes (Fig.  13b)
compared to that of ­Vs (Fig. 13c). This is because that the quick recovery of ­Vp is expected
when the strain amplitude becomes small whereas the recovery of ­Vs requires the dissipa-
tion of the excess pore water pressures.
Figure 13b shows the identified V­ p of 1,110 m/s before strong shakes. This value is con-
sistent to Fig. 2 where the averaged ­Vp in top 30 m is obtained as 1,070 m/s by assuming
the ­Vp of 300 m/s above the depth of 2.5 m. The V ­ p during strong shakes is also estimated
from the identified V­ s. When the soil shows nonlinear behaviors under unsaturated condi-
tions, the expected reduction ratio of ­Vp is the same as that of ­Vs (Tsai and Liu 2017).

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Figure 5 shows that the ­Vs decreases from 130 to 90 m/s during the strong shakes between
ground surface and the depth of 5.5 m. The reduction ratio of ­Vs is 0.69. By applying this
factor only to the V­ p between the ground surface and the depth of 2.5  m assuming the
similar reduction, the ­Vp in top 30 m is calculated during strong shakes. The results show
the ­Vp decreases from 1,070 to 950  m/s, which is consistent with the identified value of
980 m/s in Fig. 13b.
The observed reduction of ­Vp during strong shakes and its quick recovery in Fig.  13
­ p are influenced by the strain amplitude only when the soil is unsaturated.
support that the V
Based on these observations, it is concluded that the location of the unsaturated/saturated
soils is important when propagating the compression waves in vertical direction because
the nonlinear behavior of the soil is substantially different depending on these conditions
(Liu and Tsai 2018; Liu and Tsai 2020; Shi et al. 2020).

8 Conclusion

Acceleration time series and developed ­ Ru are analyzed at the Wildlife Liquefaction
Arrays. ­Vs and V ­ p are computed between sensors during strong shakes by calculating the
wave travel times using NIOM, in which V ­ s is computed at different azimuthal angles. γ
is computed by integrating the 3-component acceleration time series into displacement
time series and then dividing it with separation distances between sensors. γaz is also com-
puted at different azimuthal angles within the time window where the ­Vs is computed. The
obtained time series of V ­ s, ­Vp, and γ is compared with the ground-surface accelerations
and ­Ru at different depths. The maximum ­Ru in the analyzed data is less than 0.6.
Ru starts to accumulate when γ exceeds the threshold strain of 0.01%–0.015%. ­Vs is
negatively correlated with ­Ru and γ from within and between events. ­Vs ranges with the
azimuthal angles depending on γaz and also decreases consistently with ­Ru independent of
the azimuthal angle when γaz is limited. These observations indicate that the influence of
γaz on ­Vs variation is orientation-dependent, whereas that of ­Ru is orientation-independent.
Moreover, ­Vs and γaz seem to have a unique relationship at a given ­Ru. To the best of the
authors’ knowledge, these observations have never been previously reported. R ­ u develop-
ment is well-correlated with the cumulative maximum strain of γ and γaz, although γ is
expected to be better correlated as reported previously. This difference might be because of
the limited directionality in the analyzed ground motions.
Variations in ­Vp during strong motions were investigated and compared with those of
­Vs. ­Vp reduction is limited during strong shakes, although that of ­Vs is significant during
the development of pore water pressure. Besides, ­Vp shows the quick recovery as strain
amplitudes become small after strong shakes whereas the slow recovery of ­Vs requires the
dissipation of the excess pore-water pressures. The reduction of ­Vp in top 30 m was reason-
ably estimated during strong shakes by reducing the ­Vp above the depth of 2.5  m based
on the observed degradation of V ­ s at shallow depth but keeping the V ­ p unchanged for the
rest of depths. These observations are consistent to the recent theoretical investigation that
the reduction of constrained modulus is limited under the saturated condition because it is
determined by the combination of bulk modulus of soil skeleton and pore fluid. These find-
ings will be useful in performing 1D effective stress site response analyses by using a direc-
tional single-component ground motion under the development of pore-water pressure.

Acknowledgements The analyzed data are provided by NEES@UCSB. The Abu Dhabi Department of
Education and Knowledge (ADEK) supported the research through the 2018 ADEK Award for Research

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Excellence (AARE18-148). Access for the referenced professional journals is granted by Khalifa University
of Science and Technology, Abu Dhabi, UAE. The authors appreciate these supports for result presentation.
The second author also gratefully acknowledge the support from the Ministry of Science and Technology,
Taiwan under Award No. 109-2625-M-005 -003

Data availability  Acceleration time series and pore-water pressures are downloaded from NEES@UCSB
(http://nees.ucsb.edu/data-porta​l). Origin times, magnitudes, and hypocenter locations of analyzed earth-
quake events are obtained from Incorporated Research Institute of Seismology (IRIS) website (https​://
ds.iris.edu/ds/nodes​/dmc/data/types​/event​s/catal​ogs/).

References
Aguirre J, Irikura K (1997) Nonlinearity, liquefaction, and velocity variation of soft soil layers in Port
Island, Kobe, during the Hyogo-ken Nanbu earthquake. Bull Seismol Soc Am 87:1244–1258
Ancheta TD, Darragh RB, Stewart JP, Seyhan E, Silva WJ, Chiou BSJ, Wooddell KE, Graves RW, Kottke
AR, Boore DM, Kishida T, and Donahue JL (2013) PEER NGA-West2 database. PEER report
2013/03, Pacific Earthquake Engineering Research Center
Bennett MJ, McLaughlin PV, Sarmiento JS, Youd TL (1984) Geotechnical investigation of liquefaction
sites, imperial valley, California. U.S. Geological survey open-file report 84-252, Menlo park, Califor-
nia 94025
Bierschwale JG Stokoe KHI (1984) Analytical evaluation of liquefaction potential of sands subjected to
the 1981 Westmorland earthquake. Geotechnical engineering report GR 84-15, University of Texas,
Austin
Boore DM (2010) Orientation-independent, nongeometric-mean measures of seismic intensity
from two horizontal components of motion. Bull Seismol Soc Am 100:1830–1835. https​://doi.
org/10.1785/01200​90400​
Byrne P (1991) A cyclic shear-volume coupling and pore pressure model for sand. In: Proceedings of 2nd
international conference on recent advances in geotechnical earthquake engineering and soil
Chiou B, Darragh R, Gregor N, Silva W (2008) NGA project strong-motion database. Earthq Spectra
24:23–44
Dobry, R, Ladd, RS, Yokel, FY, Chung, RH, Powell, D (1982) Prediction of pore water pressure buildup and
liquefaction of sand by the cyclic strain method, Report, National Bureau of Standards Building, Sci-
ence Series 138, Washington, D.C., 153 p
Elgamal AW, Zeghal M, Tang HT, Stepp JC (1995) Lotung downhole seismic array. I: evaluation of site
dynamic properties. J Geotech Eng 121:350–362
Glaser SD, Baise LG (2000) System identification estimation of soil properties at the Lotung site. Soil Dyn
Earthq Eng 19:521–531
Groholski DR, Hashash YMA, Matasovic N (2014) Learning of pore pressure response and dynamic soil
behavior from downhole array measurements. Soil Dyn Earthq Eng 61–62:40–56
Haddadi HR, Kawakami H (1998) Modeling wave propagation by using normalized input-output minimiza-
tion (NIOM) method for multiple linear systems. Doboku Gakkai Ronbunshu 584:29–39
Hashash YMA, Musgrove MI, Harmon JA, Ilhan O, Xing G, Groholski DR, Phillips CA, Park D
(2020) DEEPSOIL 7.0, User manual, Urbana, IL, Board of Trustees of University of Illinois at
Urbana-Champaign
Hsu CC, Vucetic M (2004) Volumetric threshold shear strain for cyclic settlement. J Geotech Geoenviron
Eng 130:58–70
Ishihara K, Yamazaki F (1980) Cyclic simple shear tests on saturated sand in multi-directional loading.
Soils Found 20:45–59
Kishida T, Haddadi H, Darragh RB, Kayen RE, Silva WJ, Bozorgnia Y (2018) Apparent wave velocity and
site amplification at the CSMIP Carquinez bridge geotechnical arrays during the 2014 M6.0 South
Napa earthquake. Earthq Spectra EERI 34:327–347. https​://doi.org/10.1193/04231​7EQS0​74M
Kishida T, Park DS, Sousa R, Armstrong R, Byon YJ (2020) Modulus reduction of embankment dams
based on downhole array time series. Earthq Spectra 36:400–421. https​://doi.org/10.1177/87552​93019​
87818​2
Kramer SL, Hartvigsen AJ, Sideras SS, Ozener PT (2011) Site response modelling in liquefiable soil depos-
its. In: Proceedings of the 4th IASPEI/IAEE international symposium on effects of surface geology on
seismic motions. Santa Barbara, California

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Bulletin of Earthquake Engineering

Lee MKW, Finn WDL (1978) DESRA-2, dynamic effective stress response analysis of soil deposits with
energy transmitting boundary including assessment of liquefaction potential. Soil mechanics series no.
36, Department of Civil Engineering, University of British Columbia, Vancouver, Canada, 60 p
Liu HW, Tsai CC (2018) Site effect of vertical motion amplification behavior observed from downhole
arrays. J GeoEng 13:39–48
Liu PC, Tsai CC (2020) Influence of local site condition on vertical-to-horizontal spectrum ratio -Insight
from site response analysis. J Earthq Eng. https​://doi.org/10.1080/13632​469.2020.17594​73
Martin GR, Finn WDL, Seed HB (1975) Fundamentals of liquefaction under cyclic loading. J Geotech Eng
Div 101:423–438
Matasovic N, Vucetic M (1993) Seismic response of composite horizontally-layered soil deposits. UCLA
research Report No. ENG-93-182, Civil Engineering Department, University of California, Los Ange-
les, CA, March, 452 p
Matasovic N, Vucetic M (1995) Seismic response of soil deposits composed of fully-saturated clay and sand
layers. In: Ishihara K (ed) Proceedings of the 1st international conference on geotechnical earthquake
engineering, Vol. I, Tokyo, Japan. A.A. Balkema, Rotterdam/Brookfield, pp 611–616
Matasovic J, Vucetic M (1996) Analysis of seismic records from the wildlife liquefaction site. In: Sociedad
Mexicana de Ingeneria Sismica (ed) Proceedings (on CD-ROM) of the eleventh world conference on
earthquake engineering, Acapulco, Mexico. Elsevier Science Ltd., Paper No. 209, 8 pages, June
Mogi H, Kawakami E (2012) Temporal change in propagation time of seismic waves observed at vertical
array in the Onagawa Nuclear Power Plant during, before, and after the 2011 off the Pacific Coast of
Tohoku earthquake. J Jpn Assoc Earthq Eng 12:36–54 ((in Japanese))
Reyes A, Adinata J, Taiebat M (2019) Impact of bidirectional seismic shearing on the volumetric response
of sand deposits. Soil Dyn Earthq Eng 125:105665
Shi Y, Wang SY, Cheng K, Miao Y (2020) In  situ characterization of nonlinear soil behavior of vertical
ground motion using KiK-net data. Bull Earthq Eng. https​://doi.org/10.1007/s1051​8-020-00893​-1
Steidl JH, Civilini F, Seale S (2014) What have we learned after a decade of experiments and monitoring at
the NEES@UCSB permanently instrumented field sites?. In: Tenth U.S. national conference on earth-
quake Engineering frontiers of earthquake engineering, July 21–25, 2014, Anchorage, Alaska
Thilakaratne V, Vucetic M (1990) Analysis of the seismic response at the Imperial Wildlife Liquefaction
Array in 1987. In: Proceedings of fourth US national conference on earthquake engineering, May
1990, Palm Springs, California, vol. 3
Tsai CC, Hashash YMA (2008) A novel framework integrating downhole array data and site response anal-
ysis to extract dynamic soil behavior. Soil Dyn Earthq Eng 28:181–197. https​://doi.org/10.1016/j.soild​
yn.2007.06.008
Tsai CC, Hashash YMA (2009) Learning of dynamic soil behavior from downhole arrays. J Geotech Geoen-
viron Eng 135:745–757. https​://doi.org/10.1061//asce/gt.1943-5606.00000​50
Tsai CC, Liu HW (2017) Site response analysis of vertical ground motion in consideration of soil nonlinear-
ity. Soil Dyn Earthq Eng 102:124–136
Tsai CC, Chang WS, Chiou JS (2017) Enhancing prediction of ground response at the Turkey flat geotechni-
cal array. Bull Seismol Soc Am 107(5):2043–2054
Ueng TS, Lee CA (2015) Pore pressure generation in saturated sand induced by one- and two-dimensional
shakings. J GeoEng 10(2):53–61
Vucetic M, Dobry R (1986) Pore pressure build-up and liquefaction at level sandy sites during earthquakes.
Res. rep. CE-86-3, Dept. of Civil Engineering, Rensselaer Polytechnic Institute, Troy, NY
Vucetic M, Dobry R (1988) Cyclic triaxial strain-controlled testing of liquefiable sands. ASTM special tech-
nical publication 977, advanced methods for triaxial testing of soil and rock, pp. 475–485
Youd LT, Steidl J, Steller RA (2007) Instrumentation of the Wildlife Liquefaction Array. In: 4th interna-
tional conference on earthquake geotechnical engineering, Thessaloniki, Greece, June 25–28, 2007;
Paper No. 1251
Zeghal M, Elgamal AW (1994) Analysis of site liquefaction using earthquake records. J Geotech Eng ASCE
120(6):996–1017
Zeghal M, Elgamal AW, Tang HT, Stepp JC (1995) Lotung downhole array II: evaluation of soil non-linear
properties. J Geotech Eng 121(4):363–378
Zorapapel GT, Vucetic M (1994) The effects of seismic pore water pressure on ground surface motion.
Earthq Spectra Prof J Earthq Eng Res Inst (EERI) 10(2):403–437

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center
GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers
and authorised users (“Users”), for small-scale personal, non-commercial use provided that all
copyright, trade and service marks and other proprietary notices are maintained. By accessing,
sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of
use (“Terms”). For these purposes, Springer Nature considers academic use (by researchers and
students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and
conditions, a relevant site licence or a personal subscription. These Terms will prevail over any
conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription (to
the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of
the Creative Commons license used will apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may
also use these personal data internally within ResearchGate and Springer Nature and as agreed share
it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not otherwise
disclose your personal data outside the ResearchGate or the Springer Nature group of companies
unless we have your permission as detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial
use, it is important to note that Users may not:

1. use such content for the purpose of providing other users with access on a regular or large scale
basis or as a means to circumvent access control;
2. use such content where to do so would be considered a criminal or statutory offence in any
jurisdiction, or gives rise to civil liability, or is otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association
unless explicitly agreed to by Springer Nature in writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a
systematic database of Springer Nature journal content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a
product or service that creates revenue, royalties, rent or income from our content or its inclusion as
part of a paid for service or for other commercial gain. Springer Nature journal content cannot be
used for inter-library loans and librarians may not upload Springer Nature journal content on a large
scale into their, or any other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not
obligated to publish any information or content on this website and may remove it or features or
functionality at our sole discretion, at any time with or without notice. Springer Nature may revoke
this licence to you at any time and remove access to any copies of the Springer Nature journal content
which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or
guarantees to Users, either express or implied with respect to the Springer nature journal content and
all parties disclaim and waive any implied warranties or warranties imposed by law, including
merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published
by Springer Nature that may be licensed from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a
regular basis or in any other manner not expressly permitted by these Terms, please contact Springer
Nature at

onlineservice@springernature.com

You might also like